A Credo check to detect silent exception swallowing in Elixir rescue blocks.
Silent exception handling is a dangerous anti-pattern that:
- Hides bugs in production
- Makes debugging extremely difficult
- Creates unpredictable system behavior
This check enforces proper error handling by requiring that every rescue block either logs, reports to error monitoring, or re-raises the exception.
Add to your mix.exs:
def deps do
[
{:credo_exception_swallow, "~> 0.1.0", only: [:dev, :test], runtime: false}
]
endAdd to your .credo.exs in the checks: %{enabled: [...]} section:
{CredoExceptionSwallow.Checks.Warning.SilentRescue, []}{CredoExceptionSwallow.Checks.Warning.SilentRescue, [
# Exclude specific files (e.g., health checks)
files: %{excluded: ["lib/my_app_web/controllers/health_controller.ex"]},
# Set priority (:high, :normal, :low)
priority: :high,
# Skip test files (default: true)
skip_test_files: true,
# Additional acceptable function calls beyond defaults
acceptable_calls: [
"MyApp.ErrorHandler.report"
]
]}# Silent swallow - VERY BAD
try do
risky_operation()
rescue
_ -> nil
end
# Silent swallow with specific exception - STILL BAD
try do
parse_data(input)
rescue
ArgumentError -> {:error, :invalid}
end# Log the error
try do
risky_operation()
rescue
e ->
Logger.error("Operation failed: #{inspect(e)}")
{:error, :failed}
end
# Report to error monitoring
try do
risky_operation()
rescue
e ->
Sentry.capture_exception(e, stacktrace: __STACKTRACE__)
{:error, :failed}
end
# Re-raise (let it crash philosophy)
try do
risky_operation()
rescue
e -> reraise e, __STACKTRACE__
endThe following function calls are considered proper error handling:
Logger.error/1,2Logger.warning/1,2Logger.warn/1,2Logger.info/1,2Logger.debug/1,2ErrorReporter.report_exception/1,2ErrorReporter.report_message/1,2Sentry.capture_exception/1,2Sentry.capture_message/1,2reraise/2raise/1,2
MIT License - see LICENSE file.